WO2024045454A1 - 目标识别方法、存储介质及设备 - Google Patents

目标识别方法、存储介质及设备 Download PDF

Info

Publication number
WO2024045454A1
WO2024045454A1 PCT/CN2022/143573 CN2022143573W WO2024045454A1 WO 2024045454 A1 WO2024045454 A1 WO 2024045454A1 CN 2022143573 W CN2022143573 W CN 2022143573W WO 2024045454 A1 WO2024045454 A1 WO 2024045454A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
image
parameters
marker
identification
Prior art date
Application number
PCT/CN2022/143573
Other languages
English (en)
French (fr)
Inventor
赵力
Original Assignee
湖北星纪魅族科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 湖北星纪魅族科技有限公司 filed Critical 湖北星纪魅族科技有限公司
Publication of WO2024045454A1 publication Critical patent/WO2024045454A1/zh

Links

Images

Classifications

    • G06T5/80—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
    • G06T5/70—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10052—Images from lightfield camera
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20221—Image fusion; Image merging

Definitions

  • the present application belongs to the technical field of image analysis and relates to a target recognition method, and in particular to a target recognition method, storage medium and equipment.
  • wearable devices or related electronic devices such as AR (Augmented Reality), VR (Virtual Reality), smart watches, etc.
  • user identification methods are also constantly updated, and are carried out on mobile phones. Facial recognition has become relatively popular.
  • Wearable devices such as AR glasses and VR glasses can often only be worn on the head and cannot comprehensively scan facial biometric information like mobile phones.
  • a target recognition method includes: respectively acquiring a first optical image and a first depth image containing the target to be recognized, where the first depth image refers to a target containing the target to be recognized. an image of distance information; extracting a first marker image containing a first identification marker from the first optical image, and characterizing the first identification marker in the first marker image according to the first depth image
  • the parameters are corrected, wherein the first identification marker includes the first part of the target to be identified; based on the corrected first marker image, extract the target identification parameters of the first identification marker;
  • the target identification parameters are compared with pre-stored target authentication parameters to generate a target identification result.
  • the step of extracting a first marker image containing a first identification marker from the first optical image includes: inputting the first optical image into a pre-trained marker image model, An image containing the first part of the target to be identified is acquired as the first marker image.
  • the step of correcting the characteristic parameters of the first identification marker in the first marker image according to the first depth image includes: determining the first recognition marker from the first depth image.
  • the characteristic parameter of the first part in a landmark image is the first distance; obtain a second depth image of the same first part of the target to be identified in the feature recognition library, and convert the first part of the first landmark image into The distance is corrected so that it is equal to the second distance corresponding to the second image; wherein the first distance refers to the distance information of the first part in the first depth image, and the second distance refers to Distance information of the first part in the second depth image.
  • the step of comparing the target identification parameters with pre-stored target authentication parameters and generating a target identification result includes: determining that the target identification parameters are consistent with the pre-stored target authentication parameters.
  • the recognition result is successful target recognition; wherein the target recognition result includes an identity recognition result.
  • the target identification parameters include biometric parameters of the target to be identified; the step of comparing the target identification parameters with pre-stored target authentication parameters to generate a target identification result includes: determining The biometric parameters are consistent with pre-stored biometric parameters, and the target recognition result is successful target recognition.
  • the target recognition parameters also include posture characteristic parameters of the target to be recognized; the step of comparing the target recognition parameters with pre-stored target authentication parameters and generating a target recognition result includes: It is determined that the biological characteristic parameters are consistent with the pre-stored biological characteristic parameters and the posture characteristic parameters are consistent with the pre-stored posture characteristic parameters, and the target recognition result is successful target recognition.
  • the method further includes: extracting a second marker image containing a second identification marker from the first optical image, and comparing the second marker image according to the first depth image.
  • the characteristic parameters of the second recognition marker are corrected, wherein the second recognition marker includes the first item associated with the target to be recognized or the second item associated with the scene; based on the corrected second item Marker image, extract the target recognition parameter of the second recognition marker; based on the target recognition parameter of the first recognition marker and the target parameter of the second recognition marker, compare it with the pre-stored target authentication parameters , generate target recognition results.
  • the target recognition parameters of the second recognition marker are physical object characteristic parameters and posture characteristic parameters.
  • the method before the step of extracting the first identification marker from the first optical image, the method further includes: preprocessing the first optical image and the first depth image.
  • the preprocessing includes at least: image fusion and denoising processing of the first optical image and the first depth image.
  • a computer-readable storage medium is provided, a computer program is stored thereon, and the computer program implements the target identification method when executed by a processor.
  • an electronic device including: a processor and a memory; the memory stores a computer program, and the processor is configured to execute the computer program stored in the memory, so that the electronic device The device executes the target recognition method.
  • a smart wearable device which smart wearable device includes: an optical camera configured to collect an optical image containing a test target; a depth camera configured to collect a depth image containing the test target; a processor communicatively coupled to the optical camera and the depth camera; and a memory configured to store instructions that, when executed by the processor, cause the The processor executes the target recognition method.
  • FIG. 1 shows a principle flow chart of the target recognition method of the present disclosure in an embodiment.
  • FIG. 2 shows a calibration flow chart of the target recognition method in one embodiment of the present disclosure.
  • FIG. 3 shows a calibration diagram of the target recognition method of the present disclosure in one embodiment.
  • FIG. 4 shows a target recognition flow chart of the target recognition method in one embodiment of the present disclosure.
  • FIG. 5 shows a schematic structural connection diagram of the electronic device of the present disclosure in an embodiment.
  • FIG. 6 shows a schematic structural connection diagram of the smart wearable device of the present disclosure in one embodiment.
  • the target recognition method, storage medium and device provided by the embodiments of the present disclosure can combine optical images and depth images to realize the identification of different target persons and the identification of specific scenes.
  • the target recognition method provided by the embodiments of the present disclosure can be applied to electronic devices, including but not limited to mobile phones, smart wearable devices, vehicle-mounted devices, notebook computers, ultra-mobile personal computers (UMPC), netbooks, Personal digital assistant (PDA), smart speaker, TV set top box (STB) or TV, etc.
  • So-called smart wearable devices include, but are not limited to, augmented reality (AR) devices, virtual reality (VR) devices, mixed reality (Mixed Reality, MR) devices, and cinematic reality (CR) devices. wait.
  • AR augmented reality
  • VR virtual reality
  • MR mixed reality
  • CR cinematic reality
  • FIG. 1 is a schematic flow chart of the target recognition method in one embodiment of the present disclosure.
  • the target recognition method includes:
  • the first depth image refers to a depth image containing distance information (Depth) of the target to be identified.
  • the optical image may be a static image or a dynamic image of multiple frames.
  • the depth image can also be a static image or a multi-frame dynamic image.
  • using multi-frame dynamic images for recognition can achieve higher accuracy.
  • S12 Extract the first marker image containing the first identification marker from the first optical image, and correct the characteristic parameters of the first identification marker in the first marker image according to the first depth image. , wherein the first identification marker includes the first part of the target to be identified.
  • the identification markers include body parts of the target or items associated with the identity or scene.
  • the identification marker can be a part of the target's body, such as hair, face, torso, hand, arm or leg, or other body parts, or it can be a specific item, such as a pen holder in the office, a pen in the hand, etc.
  • the characteristics expressed by the target's body parts can be used to identify his or her identity
  • the characteristics expressed by the items associated with the identity can be used to identify a certain person's identity
  • the characteristics expressed by the items associated with the scene can be used to identify a certain person. specific scenario.
  • the target person may be the device wearer or a target to be identified other than the device wearer.
  • step S12 includes:
  • the first optical image is input to a pre-trained landmark image model to determine whether the identification landmark is at least one of a body part of the target, an item associated with an identity, or an item associated with a scene.
  • the target body part image is a palm image.
  • the landmark image model is a machine learning model.
  • the machine learning model can use sample images and their annotations to train the initial neural network and evaluate, adjust, and optimize the parameters of the neural network through the loss function to achieve the preset conditions of the loss function. , for example, the loss function converges, the number of iterations reaches the preset number, or the loss value of the loss function is less than the preset threshold, thereby obtaining a neural network that meets the needs.
  • S122 Determine the characteristic parameter of the first part in the first landmark image as the first distance from the first depth image.
  • the first part of the first landmark image is determined based on the first depth image.
  • the first part is located at the lowermost end, the uppermost end of the target to be identified, or any other part that can be determined based on the depth image; the characteristic parameters of the first part include the first distance.
  • FIG. 3 is a calibration diagram of the target recognition method of the present application in one embodiment.
  • Figure 3 is a simple example of optical image data conversion, in which the gesture stored in the device feature recognition library is the first palm image obtained by the device user at a distance of 50cm, and the first palm image is vertical, and The gesture obtained during the recognition process is the second palm image at a distance of 70cm. At the same time, the second palm image has a certain inclination angle relative to the first palm image. Then, through geometric principles such as the cosine theorem, the second palm image can be corrected so that the corrected second palm image has the same direction and size as the first palm image.
  • step S12 may also be to first correct the characteristic parameters of the identification markers based on the depth image, and then extract the identification markers from the optical image.
  • S14 Compare the target recognition parameters with pre-stored target authentication parameters to generate a target recognition result; wherein the target recognition result includes an identity recognition result.
  • the target recognition parameters are consistent with the pre-stored target authentication parameters, and the target recognition result is successful target recognition.
  • the target recognition result is target recognition failure.
  • the target recognition results include identity recognition results and/or scene recognition results.
  • the target to be recognized is five fingers, and the length error of the five fingers is set to be within 2mm, then when the length errors of the five fingers in the target recognition parameters are all within 2mm, the five-finger target is successfully recognized; when there is one or more finger length errors If it exceeds 2mm, the target recognition of the five fingers will fail.
  • the threshold range (for example, within 2 mm) can be set according to the resolution of the depth camera. The higher the hardware resolution of the depth camera, the smaller the threshold range can be set.
  • the target identification parameters include the biometric parameters of the target to be identified; the step of comparing the target identification parameters with pre-stored target authentication parameters and generating a target identification result includes:
  • biometric parameters are consistent with the pre-stored biometric parameters, and the target recognition result is successful target recognition;
  • the target recognition result is target recognition failure.
  • the target's biometric parameters can be physical parameters of the target's body, such as the length of the five fingers, the length of the index finger knuckles, the width of the fingers, the arm, the location of specific moles or scars on the arm, and the length of the forearm. , characteristics of palm prints and birthmarks of specific sizes, etc.; biometric parameters of specific items are also physical characteristic parameters, such as the size, shape, color, etc. of the item.
  • the target recognition parameters further include posture characteristic parameters of the target to be recognized; the step of comparing the target recognition parameters with pre-stored target authentication parameters to generate a target recognition result includes:
  • the biological characteristic parameters are consistent with the pre-stored biological characteristic parameters and the posture characteristic parameters are consistent with the pre-stored posture characteristic parameters, and the target recognition result is successful target recognition.
  • the target recognition result is target recognition failure.
  • the target's posture characteristic parameters can be the specific posture of the target's body during the recognition process, such as the angle between the palm and the forearm, the OK gesture made by the left hand, crossing the arms across the chest, etc.; specific items
  • the posture characteristic parameters can be its placement position, inclination angle, height, distance from the background wall, etc.
  • the method further includes: extracting a second marker image including a second identification marker from the first optical image, and comparing the second marker image according to the first depth image.
  • the characteristic parameters of the second recognition marker are corrected, wherein the second recognition marker includes the first item associated with the target to be recognized or the second item associated with the scene; based on the corrected second item Marker image, extract the target recognition parameter of the second recognition marker; based on the target recognition parameter of the first recognition marker and the target parameter of the second recognition marker, compare it with the pre-stored target authentication parameters , generate target recognition results.
  • the target recognition parameters of the second recognition marker are physical feature parameters and posture feature parameters.
  • the step of comparing the target identification parameters with pre-stored target authentication parameters to generate a target identification result includes:
  • biometric parameters can be used for recognition
  • more than two posture feature parameters can be used for recognition
  • one or more A biological characteristic parameter is combined with one or more posture characteristic parameters for identification.
  • the accuracy of recognition is improved by increasing the complexity of recognition. For example, identifying the hand characteristics of the target person and the pen dedicated to the user in the study room, etc.
  • For scene recognition for example: 1. Hang a specific painting or place a cabinet in the conference room scene. After the electronic device used by the user executes the target recognition method, for example, after scanning, it recognizes that the current scene is a conference room and automatically accesses the scene. s meeting. 2. In an exhibition scenario, after the electronic device used by the user performs the target recognition method, for example, after scanning a specific product, the relevant product information corresponding to the exhibition scenario is identified, and the exchange of personal information and manufacturer information of the exhibition products is further completed through the presented interface. . 3.
  • the electronic device used by the user In a scene outside a user's car, after the electronic device used by the user performs the target recognition method, for example, scanning from outside the car to a specific pendant in the car, it identifies the user's vehicle, thereby remotely activating or deactivating the vehicle or other specific functions. . 4.
  • the target recognition method such as scanning from outside the car to a specific pendant in the car
  • the call function of the AR glasses is turned off, and the call function is switched to the in-vehicle device.
  • the external human-computer interaction interface set or presented through the software program can be flexibly set for different logo identifiers and characteristic parameters to achieve limitations on users and usage scenarios.
  • a dialog box presented through a human-computer interaction interface For example, a dialog box presented through a human-computer interaction interface.
  • biometric parameters multiple biometric parameters form a biometric parameter list
  • multiple posture feature parameters form a posture feature parameter list.
  • Biometric parameters and posture feature parameters can be automatically configured through active guidance, such as presenting a list of biometric parameters for the user to select, or they can be set by themselves, such as the user can create a new gesture.
  • the target recognition method further includes:
  • the first optical image and the first depth image are preprocessed, and the preprocessing includes at least: image fusion and denoising of the first optical image and the first depth image.
  • image fusion refers to correlating the distance data of the depth image on the optical image, thereby increasing the data dimension of the optical image.
  • FIG. 4 shows a target recognition flow chart of the target recognition method of the present application in one embodiment.
  • the target recognition process includes:
  • S52 Extract the image data (optical image and depth image data) of the identification marker.
  • S54 Extract the biometric parameters and posture feature parameters of the feature marker from the corrected recognition marker image.
  • S55 Compare the biometric parameters and posture feature parameters with the identity authentication information stored in the device. If the biometric parameters and posture feature parameters match the identity authentication information stored in the device, it is determined that the identity recognition is successful; if the biometric parameters are consistent with the identity authentication information stored in the device, it is determined that the identity recognition is successful; If the parameters and posture feature parameters are inconsistent with the identity authentication information stored in the device, it is determined that the identity recognition has failed.
  • the protection scope of the target identification method described in the present disclosure is not limited to the execution sequence of the steps listed in the embodiments. All solutions implemented by adding or subtracting steps or replacing steps based on the principles of this application are included in the protection scope of this application.
  • Embodiments of the present disclosure provide a computer-readable storage medium on which a computer program is stored.
  • the computer program implements the target identification method when executed by a processor.
  • the aforementioned computer program can be stored in a computer-readable storage medium.
  • the steps including the above-mentioned method embodiments are executed; and the aforementioned computer-readable storage media include: ROM, RAM, magnetic disks, optical disks and other computer storage media that can store program codes.
  • FIG. 5 is a schematic structural connection diagram of the electronic device of the present application in one embodiment.
  • an embodiment of the present disclosure provides an electronic device 6, including: a processor 61 and a memory 62; the memory 62 stores a computer program, and the processor 61 is configured to execute the memory 62 to store a computer program.
  • the processor 61 may be, for example, a central processing unit (Central Processing Unit, referred to as CPU); it may also be a graphics processor (Graphics Processing Unit, referred to as GPU), a digital signal processor (Digital Signal Processing, referred to as DSP), a dedicated Integrated circuit (Application Specific Integrated Circuit, ASIC for short), Field Programmable Gate Array (FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • CPU Central Processing Unit
  • GPU Graphics Processing Unit
  • DSP Digital Signal Processing
  • ASIC Application Specific Integrated Circuit
  • FPGA Field Programmable Gate Array
  • the above-mentioned memory 62 may include random access memory (Random Access Memory, RAM for short), and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
  • RAM Random Access Memory
  • non-volatile memory such as at least one disk memory.
  • the electronic device may include a memory, a processor, a peripheral interface, an RF circuit, an audio circuit, a speaker, a microphone, an input/output (I/O) subsystem, a display screen, and other output or control devices.
  • the computers include but are not limited to personal computers such as desktop computers, laptop computers, tablet computers, smartphones, smart TVs, personal digital assistants (Personal Digital Assistant, or PDA),
  • PDA Personal Digital Assistant
  • the electronic device may also be a vehicle terminal.
  • the electronic device may also be a server.
  • the server may be arranged on one or more physical servers according to various factors such as function and load, or may be composed of a distributed or centralized server cluster.
  • the cloud server is not limited in this embodiment.
  • the electronic device includes a smart wearable device 7 (for example, an electronic device such as AR, VR, MR, CR, etc.).
  • the smart wearable device 7 includes:
  • Optical camera 71 is configured to collect an optical image including a test target, where the test target includes a target to be identified, and the optical image includes a first optical image; in one embodiment, the optical camera includes an optical lens; a motor , such as stepper motor, ultrasonic motor, voice coil motor (VCM, Voice Circle Motor/Voice Coil Actuator), etc.; photosensitive chip, such as CCD or CMOS; drive circuit; output interface and other components .
  • a motor such as stepper motor, ultrasonic motor, voice coil motor (VCM, Voice Circle Motor/Voice Coil Actuator), etc.
  • VCM Voice Circle Motor/Voice Coil Actuator
  • the depth camera 72 is configured to collect a depth image containing a test target, where the depth image includes a first depth image; the depth camera can, for example, use structured light (Structured-light), binocular vision (Stereo), or light time-of-flight method. (TOF) and so on to achieve depth image acquisition.
  • structured-light Structured-light
  • Stepo binocular vision
  • TOF light time-of-flight method.
  • the depth camera is a TOF camera, which includes an illumination unit including a light source and a pulse modulator; an optical lens provided with a bandpass filter; a TOF dedicated photosensitive chip; and a control unit that controls the pulse and chip on/off Closed precise synchronization; computing unit used to record an accurate depth map, where the depth map is a grayscale image, where each value represents the distance between a light-reflecting surface and the camera.
  • Processor 73 in communication with the optical camera and the depth camera; and memory 74, the memory is configured to store instructions that, when the stored instructions are executed by the processor, cause the processor to perform steps, The steps include:
  • the first depth image refers to an image containing distance information of the target to be identified
  • a first marker image containing a first identification marker is extracted from the first optical image, and the characteristic parameters of the first identification marker in the first marker image are corrected according to the first depth image, wherein , the first identification marker includes the first part of the target to be identified;
  • the target identification parameters are compared with pre-stored target authentication parameters to generate a target identification result.
  • AR products are widely used in smart homes, smart transportation, and assisted driving. It can be used with AR devices and has high playability. For example, in terms of smart homes, AR is used to recognize objects at home, and when you get home, AR can be used to recognize specific gestures to start furniture and equipment; in terms of smart transportation, AR is used to browse traffic information; in terms of assisted driving, drivers wear AR glasses for navigation, etc. Therefore, compared with fingerprint recognition, through optical cameras and TOF cameras, there is no need to touch the device every time for recognition.
  • the target recognition method, storage medium and device provided by the embodiments of the present disclosure have at least partially the following beneficial effects:
  • data processing and recognition are performed to realize the identification of the identity of the device wearer or user. It can be applied to wearable devices such as AR and VR, and has application prospects in many fields such as smart homes, smart transportation, and assisted driving.
  • Utilizing a depth camera can provide distance data that complements the image data from an optical camera.
  • the depth camera is used to measure the target size, and the optical camera is used to obtain more target details. Combining the two can be used for identity recognition and scene recognition.
  • Contactless recognition can be achieved; the accuracy and robustness of recognition can be improved by combining biometric features with posture features such as gestures; each recognition feature can be flexibly set, with high configurability.

Abstract

本申请提供一种目标识别方法、存储介质及设备，所述目标识别方法包括：获取包含待识别目标的第一光学图像和第一深度图像，所述第一深度图像是指包含所述待识别目标距离信息的图像；从所述第一光学图像中提取包含第一识别标志物的第一标志物图像，根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正；基于校正后的所述第一标志物图像，提取所述第一识别标志物的目标识别参数；将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果。本申请将光学图像和深度图像结合，实现不同目标者身份的识别以及特定场景的识别。

Description

目标识别方法、存储介质及设备
本申请要求于2022年8月29日提交中国专利局、申请号为2022110474634、申请名称为“目标识别方法、存储介质及设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请属于图像分析的技术领域，涉及一种目标识别方法，特别是涉及一种目标识别方法、存储介质及设备。
背景技术
随着AR(Augmented Reality，增强现实)、VR(Virtual Reality，虚拟现实)、智能手表等可穿戴设备或相关电子设备的普及及发展，使用者的身份识别方法也在不断更新，在手机中进行人脸识别已经相对普及，AR眼镜、VR眼镜等可穿戴设备，往往只能佩戴在头部，无法像手机一样实现对人脸生物信息的全面扫描。
发明内容
在本公开的一个方面，提供一种目标识别方法，所述方法包括：分别获取包含待识别目标的第一光学图像和第一深度图像，所述第一深度图像是指包含所述待识别目标距离信息的图像；从所述第一光学图像中提取包含第一识别标志物的第一标志物图像，根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正，其中，所述第一识别标志物包括所述待识别目标的第一部位；基于校正后的所述第一标志物图像，提取所述第一识别标志物的目标识别参数；将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果。
在一些实施例中，所述从所述第一光学图像中提取包含第一识别标志物的第一标志物图像步骤，包括：将所述第一光学图像输入至预先训练的标志物图像模型，获取包含所述待识别目标的第一部位的图像作为所述第一标志物图像。
在一些实施例中，所述根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正的步骤，包括：由所述第一深度图像确定所述第一标志物图像中所述第一部位的特征参数为第一距离；获取特征识别库中同一所述待识别目标第一部位的第二深度图像，将所述第一标志物图像中的第一距离校正，使其与所述第二图像对应的第二距离相等； 其中，所述第一距离是指所述第一深度图像中所述第一部位的距离信息，所述第二距离是指所述第二深度图像中所述第一部位的距离信息。
在一些实施例中，所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：确定所述目标识别参数与预存的目标认证参数相符，所述目标识别结果为目标识别成功；其中，所述目标识别结果包括身份识别结果。
在一些实施例中，所述目标识别参数包括所述待识别目标的生物特征参数；所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：确定所述生物特征参数与预存的生物特征参数相符，所述目标识别结果为目标识别成功。
在一些实施例中，所述目标识别参数还包括所述待识别目标的姿态特征参数；所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：确定所述生物特征参数与预存的生物特征参数相符且所述姿态特征参数与预存的姿态特征参数相符，所述目标识别结果为目标识别成功。
在一些实施例中，所述方法还包括：从所述第一光学图像中提取包含第二识别标志物的第二标志物图像，根据所述第一深度图像对所述第二标志物图像中第二识别标志物的特征参数进行校正，其中，所述第二识别标志物包括所述待识别目标相关联的第一物品或与场景相关联的第二物品；基于校正后的所述第二标志物图像，提取所述第二识别标志物的目标识别参数；基于所述第一识别标志物的目标识别参数和所述第二识别标志物的目标参数，与预存的目标认证参数进行比对，生成目标识别结果。
在一些实施例中，所述第二识别标志物的目标识别参数为实物特征参数和姿态特征参数。
在一些实施例中，在所述从所述第一光学图像中提取第一识别标志物的步骤之前，所述方法还包括：对所述第一光学图像和所述第一深度图像进行预处理，所述预处理至少包括：所述第一光学图像和所述第一深度图像的图像融合以及去噪处理。
在本公开的另一方面，提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现所述的目标识别方法。
在本公开的又一个方面，提供一种电子设备，包括：处理器及存储器；所述存储器存储有计算机程序，所述处理器被配置成执行所述存储器存储的计算机程序，以使所述电子设备执行所述的目标识别方法。
在本公开的还一个方面，提供一种智能穿戴设备，所述智能穿戴设备包括：光学摄像头，被配置成采集包含测试目标的光学图像；深度摄像头，被配置成采集包含测试目标的深度图像；处理器，所述处理器与所述光学摄像头、所述深度摄像头可通信地耦合；以及存储器， 所述存储器被配置成存储指令，当所述存储指令被所述处理器执行时，使所述处理器执行所述的目标识别方法。
附图说明
图1显示为本公开的目标识别方法于一实施例中的原理流程图。
图2显示为本公开的目标识别方法于一实施例中的校正流程图。
图3显示为本公开的目标识别方法于一实施例中的校正示意图。
图4显示为本公开的目标识别方法于一实施例中的目标识别流程图。
图5显示为本公开的电子设备于一实施例中的结构连接示意图。
图6显示为本公开的智能穿戴设备于一实施例中的结构连接示意图。
具体实施方式
以下通过特定的具体实例说明本申请的实施方式，本领域技术人员可由本说明书所揭露的内容轻易地了解本申请的其他优点与功效。本申请还可以通过另外不同的具体实施方式加以实施或应用，本说明书中的各项细节也可以基于不同观点与应用，在没有背离本申请的精神下进行各种修饰或改变。需说明的是，在不冲突的情况下，以下实施例及实施例中的特征可以相互组合。
需要说明的是，以下实施例中所提供的图示仅以示意方式说明本申请的基本构想，遂图示中仅显示与本申请中有关的组件而非按照实际实施时的组件数目、形状及尺寸绘制，其实际实施时各组件的型态、数量及比例可为一种随意的改变，且其组件布局型态也可能更为复杂。
本公开实施例提供的目标识别方法、存储介质及设备可以将光学图像和深度图像结合，实现不同目标者身份的识别以及特定场景的识别。
本公开实施例提供的目标识别方法可以应用于电子设备上，电子设备包括但不限于手机、智能穿戴设备、车载设备、笔记本电脑、超级移动个人计算机(ultra-mobile personal computer，UMPC)、上网本、个人数字助理(personal digital assistant，PDA)、智能音箱、电视机顶盒(set top box，STB)或电视等。所称的智能穿戴设备，包括但不限于增强现实(augmented reality，AR)设备、虚拟现实(virtual reality，VR)设备、混合现实(Mixed Reality，MR)设备、影像现实(Cinematic Reality，CR)设备等。
以下将结合图1至图6详细阐述本实施例的一种目标识别方法、存储介质及设备的原理及实施方式。
请参阅图1，显示为本公开的目标识别方法于一实施例中的原理流程图。如图1所示，所述目标识别方法包括：
S11，分别获取包含待识别目标的第一光学图像和第一深度图像，所述第一深度图像是指包含所述待识别目标的距离信息(Depth)的深度图像。
例如，光学图像可以是静态图像或者多帧的动态图像。深度图像也可以是静态图像或者多帧的动态图像。
例如，利用多帧的动态图像进行识别，准确性更高。
S12，从所述第一光学图像中提取包含第一识别标志物的第一标志物图像，根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正，其中，所述第一识别标志物包括所述待识别目标的第一部位。
于一实施例中，所述识别标志物包括目标者身体部位或与身份、场景关联的物品。具体地，识别标志物可以是目标者的身体的一部分，例如头发、脸部、躯干、手、手臂或腿等身体部位，也可以是特定的物品，比如办公室的笔筒，手中的钢笔等。其中，目标者身体部位所表现的特征可以用于标识其身份，与身份关联的物品所表现的特征可以用于标识某个人的身份，与场景关联的物品所表现的特征可以用于标识某个特定的场景。其中，目标者可以是设备佩戴者，也可以是设备佩戴者以外的待识别目标。
请参阅图2，显示为本申请的目标识别方法于一实施例中的校正流程图。如图2所示，步骤S12包括：
S121，将所述第一光学图像输入至预先训练的标志物图像模型，获取包含所述待识别目标的第一部位的图像作为所述第一标志物图像。
例如，将第一光学图像输入至预先训练好的标志物图像模型，以确定所述识别标志物是否为目标者身体部位、与身份关联的物品或与场景关联的物品中的至少一种。
于一实施例中，所述目标者身体部位图像为手掌图像。
例如，所述标志物图像模型为一机器学习模型。尽管没有明确描述的，本领域技术人员能够理解，机器学习模型可以利用样本图像及其标注对初始神经网络进行训练形成，通过损失函数评估、调整、优化神经网络的参数使损失函数达到预设条件，例如损失函数收敛、迭代次数达到预设次数或者损失函数的损失值小于预设阈值，从而得到符合需求的神经网络。
S122，由所述第一深度图像确定所述第一标志物图像中所述第一部位的特征参数为第一距离。
例如，根据所述第一深度图像确定所述第一标志图像的所述第一部位。
在一个实施例中，所述第一部位位于所述待识别目标的最下端，最上端，或任何其他可以根据深度图像确定的部位；所述第一部位的特征参数包含第一距离。
S123，获取特征识别库中同一所述待识别目标第一部位的第二深度图像，将所述第一标志物图像中的第一距离校正，使其与所述第二图像对应的第二距离相等；其中，所述第一距离是指所述第一深度图像所包含的所述第一部位的距离信息，所述第二距离是指所述第二深度图像所包含的所述第一部位的距离信息。
在一个实施例中，根据同一手掌的两个深度图像，将所述第一距离校正为特征识别库中与所述手掌图像对应的第二距离；其中，所述第一距离和第二距离是指智能穿戴设备获取所述手掌图像时的拍摄点与所述手掌之间的距离，即图3中的S i，i＝1,2,3,4…。
请参阅图3，显示为本申请的目标识别方法于一实施例中的校正示意图。图3是一个简单的光学图像数据换算的示例，其中，设备特征识别库里存储的手势是设备使用者在50cm的距离上获取的第一手掌图像，且第一手掌图像是竖直的，而识别过程中获取的手势是70cm距离上的第二手掌图像，同时第二手掌图像相对于第一手掌图像存在一定倾角。则通过余弦定理等几何原理，可将第二手掌图像校正，使校正的第二手掌图像后与第一手掌图像方向、大小一致。
于另一实施例中，步骤S12也可以是先根据深度图像对所述识别标志物的特征参数进行校正，然后再从光学图像中提取识别标志物。
S13，基于校正后的所述第一标志物图像，提取所述第一识别标志物的目标识别参数。
S14，将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果；其中，所述目标识别结果包括身份识别结果。
例如，确定所述目标识别参数与预存的目标认证参数相符，所述目标识别结果为目标识别成功。
容易理解，所述目标识别参数与预存的目标认证参数不相符，则所述目标识别结果为目标识别失败。
例如，所述目标识别结果包括身份识别结果和/或场景识别结果。
例如，待识别目标为五指，且设置五指长度误差需在2mm以内，则当所述目标识别参数中五指长度误差都在2mm以内，五指这一目标识别成功；当存在一个或多个手指长度误差超出2mm，则五指这一目标识别失败。其中，所述阈值范围(比如，2mm以内)可根据深度摄像头的分辨率进行设置，深度摄像头的硬件分辨率越高，阈值范围可以设置得越小。
于一实施例中，所述目标识别参数包括所述待识别目标的生物特征参数；所述将所述目 标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：
确定所述生物特征参数与预存的生物特征参数相符，所述目标识别结果为目标识别成功；
容易理解，所述生物特征参数与预存的生物特征参数不相符，则所述目标识别结果为目标识别失败。
例如，目标者的生物特征参数可以是目标者身体的物理参数，比如五根手指的长度、食指指节的长度、手指的宽度、手臂、手臂上特定痣或疤痕的位置、胳膊小臂的长度、掌纹的特征以及特定大小的胎记等；特定物品的生物特征参数也就是实物特征参数，比如物品的大小、形状、颜色等。
于一实施例中，所述目标识别参数还包括所述待识别目标的姿态特征参数；所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：
确定所述生物特征参数与预存的生物特征参数相符且所述姿态特征参数与预存的姿态特征参数相符，所述目标识别结果为目标识别成功。
容易理解，所述生物特征参数与预存的生物特征参数不相符或所述姿态特征参数与预存的姿态特征参数不相符，则所述目标识别结果为目标识别失败。
例如，目标者的姿态特征参数可以是目标者在识别过程中身体呈现的特定姿势，比如手掌撑起后和小臂的夹角，左手比出的OK手势，双臂交叉抱胸等；特定物品的姿态特征参数可以是其摆放的位置，倾角，高度，与背景墙的距离等。
于一实施例中，所述方法还包括：从所述第一光学图像中提取包含第二识别标志物的第二标志物图像，根据所述第一深度图像对所述第二标志物图像中第二识别标志物的特征参数进行校正，其中，所述第二识别标志物包括所述待识别目标相关联的第一物品或与场景相关联的第二物品；基于校正后的所述第二标志物图像，提取所述第二识别标志物的目标识别参数；基于所述第一识别标志物的目标识别参数和所述第二识别标志物的目标参数，与预存的目标认证参数进行比对，生成目标识别结果。
于一实施例中，所述第二识别标志物的目标识别参数为实物特征参数和姿态特征参数。
于又一实施例中，所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：
确定存在至少两种所述目标识别参数与预存的目标认证参数相符，所述目标识别结果为目标识别成功。
具体地，为了进一步提高识别的准确性，针对身份识别或场景识别，可以采用两种以上的生物特征参数进行识别，也可以采用两种以上的姿态特征参数进行识别，还可以是一种或 多种生物特征参数与一种或多种姿态特征参数结合进行识别。以此，通过增加识别的复杂度，提高识别的准确性。比如识别目标者的手部特征加上书房里该使用者专用的笔等。
针对场景识别，例如：1.会议室场景里挂一个特定的画或者放一个柜子，用户所用的电子设备执行目标识别方法后，例如扫描后识别出当前为会议室场景，自动接入该场景下的会议。2.展会场景下，用户所用的电子设备执行目标识别方法后，例如扫描特定产品后，识别出该展会场景对应的相关产品信息，进一步通过呈现的界面完成个人信息与展会产品的厂商信息的交换。3.某用户车外场景下，该用户所用的电子设备执行目标识别方法后，例如从车外扫描到车内特定挂件，识别出为该用户车辆，由此远程激活或关闭车辆或者其他特定功能。4.某用户车内场景下，该用户所用的电子设备执行目标识别方法后，例如从车外扫描到车内特定挂件，则关闭AR眼镜的通话功能，通话功能切换到车载设备。
在一些实施里中，通过软件程序设置或呈现的对外的人机交互界面，可以针对不同的标志识别物和特征参数进行灵活设置，以实现使用者和使用场景的限定。例如，通过人机交互界面呈现的对话框。于实际应用中，针对生物特征参数，多个生物特征参数形成生物特征参数列表，多个姿态特征参数形成姿态特征参数列表。生物特征参数和姿态特征参数可以是通过主动引导自动配置的，例如呈现生物特征参数列表供用户选择，也可以是自己设置的，例如用户可以创建新的手势。
于一实施例中，在步骤S11之后，在步骤S12之前，所述目标识别方法还包括：
对所述第一光学图像和所述第一深度图像进行预处理，所述预处理至少包括：所述第一光学图像和所述第一深度图像的图像融合以及去噪处理。其中，图像融合是指是在光学图像上关联深度图像的距离数据，以此增加光学图像部分数据维度。
请参阅图4，显示为本申请的目标识别方法于一实施例中的目标识别流程图。如图4所示，目标识别的流程包括：
S51，输入光学摄像头和深度摄像头数据，进行图像融合等预处理。
S52，提取识别标志物的图像数据(光学图像和深度图像数据)。
S53，根据深度摄像头数据，识别标志物的光学图像进行校正。
S54，从校正后的识别标志物图像中提取特征标志物的生物特征参数和姿态特征参数。
S55，将生物特征参数和姿态特征参数与设备中存储的身份认证信息进行比对，若生物特征参数和姿态特征参数与设备中存储的身份认证信息符合，则判定为身份识别成功；若生物特征参数和姿态特征参数与设备中存储的身份认证信息不符，则判定为身份识别失败。
本公开所述的目标识别方法的保护范围不限于实施例列举的步骤执行顺序，凡是根据本 申请的原理所做的步骤增减、步骤替换所实现的方案都包括在本申请的保护范围内。
本公开的实施例提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现所述目标识别方法。
本领域普通技术人员可以理解：实现上述各方法实施例的全部或部分步骤可以通过计算机程序相关的硬件来完成。前述的计算机程序可以存储于一计算机可读存储介质中。该程序在执行时，执行包括上述各方法实施例的步骤；而前述的计算机可读存储介质包括：ROM、RAM、磁碟或者光盘等各种可以存储程序代码的计算机存储介质。
请参阅图5，显示为本申请的电子设备于一实施例中的结构连接示意图。如图5所示，本公开的实施例提供一种电子设备6，包括：处理器61及存储器62；所述存储器62存储有计算机程序，所述处理器61被配置成执行所述存储器62存储的计算机程序，以使所述电子设备6执行前述任意实施例所述目标识别方法的一个或多个步骤。
所述处理器61，例如可以是中央处理器(Central Processing Unit，简称CPU)；还可以是图形处理器(Graphics Processing Unit，简称GPU)、数字信号处理器(Digital Signal Processing，简称DSP)、专用集成电路(Application Specific Integrated Circuit，简称ASIC)、现场可编程门阵列(FieldProgrammable GateArray，简称FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
上述的存储器62可能包含随机存取存储器(Random Access Memory，简称RAM)，也可能还包括非易失性存储器(non-volatile memory)，例如至少一个磁盘存储器。
于实际应用中，所述电子设备可以是包括存储器、处理器、外设接口、RF电路、音频电路、扬声器、麦克风、输入/输出(I/O)子系统、显示屏、其他输出或控制设备，以及外部端口等所有或部分组件的计算机；所述计算机包括但不限于如台式电脑、笔记本电脑、平板电脑、智能手机、智能电视、个人数字助理(Personal Digital Assistant，简称PDA)等个人电脑，所述电子设备还可以是车机端。在另一些实施方式中，所述电子设备还可以是服务器，所述服务器可以根据功能、负载等多种因素布置在一个或多个实体服务器上，也可以是由分布的或集中的服务器集群构成的云服务器，本实施例不作限定。
在本公开的一些实施例中，所述电子设备包括智能穿戴设备7(例如，AR、VR、MR、CR等电子设备)。参考图6，所述智能穿戴设备7包括：
光学摄像头71，被配置成采集包含测试目标的光学图像，其中所述测试目标包含待识别目标，所述光学图像包含第一光学图像；在一实施例中，所述光学摄像头包括光学透镜；马达，如步进马达(Stepper Motor)、超声波马达(Ultrasonic motor)、音圈马达(VCM，Voice  Circle Motor/Voice Coil Actuator)等；感光芯片，如CCD或CMOS；驱动电路；输出接口等多种元件。
深度摄像头72，被配置成采集包含测试目标的深度图像，其中所述深度图像包含第一深度图像；深度摄像头例如可以通过结构光(Structured-light)、双目视觉(Stereo)、光飞行时间法(TOF)等实现深度图像采集。
在一实施例中，所述深度摄像头为TOF摄像头，包括照射单元，包含光源和脉冲调制器；光学透镜，设置有带通滤光片；TOF专用感光芯片；控制单元，控制脉冲与芯片开/闭精确同步；计算单元，用于记录精确的深度图，其中深度图为灰度图，其中的每个值代表一个光反射表面和相机之间的距离。
处理器73，与所述光学摄像头、所述深度摄像头通信；以及存储器74，所述存储器被配置成存储指令，当所述存储指令被所述处理器执行时，使所述处理器执行步骤，所述步骤包括：
调用由所述光学摄像头采集的所述第一光学图像，以及由所述深度摄像头采集的所述第一深度图像，所述第一深度图像是指包含所述待识别目标的距离信息的图像；
从所述第一光学图像中提取包含第一识别标志物的第一标志物图像，根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正，其中，所述第一识别标志物包括所述待识别目标的第一部位；
基于校正后的所述第一标志物图像，提取所述第一识别标志物的目标识别参数；
将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果。
本公开实施例提供的技术方案，可以应用于AR、VR、MR等智能眼镜产品、智能手表及其他智能穿戴设备上进行身份识别和场景识别，AR类产品在智能家居、智慧交通、辅助驾驶都可以搭配AR类设备使用，可玩性高。比如智能家居方面家庭里用AR识物，到家后，可通过AR识别特定手势进行家具设备的启动等；智慧交通方面利用AR浏览交通信息；辅助驾驶方面驾驶者佩戴AR眼镜进行导航等。由此，与指纹的识别相比，通过光学摄像头和TOF摄像头，不需要每次识别时都触摸设备。
本公开实施例提供的目标识别方法、存储介质及设备，至少部分地具有以下有益效果：
基于深度摄像头结合光学摄像头采集静态图像，或多帧的动态图像，进行数据处理和识别，实现设备佩戴者或使用者身份的识别。可以应用于AR、VR等可穿戴设备，在智能家居、智慧交通、辅助驾驶等多领域均有应用前景。
利用深度摄像头可以提供距离数据，作为光学摄像头的图像数据的有力补充。通过深度 摄像头实现目标物尺寸的测量，通过光学摄像头获取更多的目标物细节，将两者相结合，可以用于身份识别和场景识别。
可以实现无接触识别；通过生物特征结合手势等姿态特征，提高识别的准确性和鲁棒性；可灵活设置各个识别特征，可配置性高。
上述实施例仅例示性说明本申请的原理及其功效，而非用于限制本申请。任何熟悉此技术的人士皆可在不违背本申请的精神及范畴下，对上述实施例进行修饰或改变。因此，举凡所属技术领域中具有通常知识者在未脱离本申请所揭示的精神与技术思想下所完成的一切等效修饰或改变，仍应由本申请的权利要求所涵盖。

Claims (12)

  1. 一种目标识别方法，其中，所述方法包括：
    分别获取包含待识别目标的第一光学图像和第一深度图像，所述第一深度图像包含所述待识别目标距离信息的图像；
    从所述第一光学图像中提取包含第一识别标志物的第一标志物图像，根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正，其中，所述第一识别标志物包括所述待识别目标的第一部位；
    基于校正后的所述第一标志物图像，提取所述第一识别标志物的目标识别参数；
    将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果。
  2. 根据权利要求1所述的目标识别方法，其中，所述从所述第一光学图像中提取包含第一识别标志物的第一标志物图像步骤，包括：
    将所述第一光学图像输入至预先训练的标志物图像模型，获取包含所述待识别目标的第一部位的图像作为所述第一标志物图像。
  3. 根据权利要求2所述的目标识别方法，其中，所述根据所述第一深度图像对所述第一标志物图像中第一识别标志物的特征参数进行校正的步骤，包括：
    由所述第一深度图像确定所述第一标志物图像中所述第一部位的特征参数为第一距离；
    获取特征识别库中同一所述待识别目标第一部位的第二深度图像，将所述第一标志物图像中的第一距离校正,使其与所述第二图像对应的第二距离相等；
    其中，所述第一距离是指所述第一深度图像中所述第一部位的距离信息，
    所述第二距离是指所述第二深度图像中所述第一部位的距离信息。
  4. 根据权利要求1所述的目标识别方法，其中，所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：
    确定所述目标识别参数与预存的目标认证参数相符，所述目标识别结果为目标识别成功；
    其中，所述目标识别结果包括身份识别结果。
  5. 根据权利要求4所述的目标识别方法，其中，所述目标识别参数包括所述待识别目标的生物特征参数；所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：
    确定所述生物特征参数与预存的生物特征参数相符，所述目标识别结果为目标识别成功。
  6. 根据权利要求5所述的目标识别方法，其中，所述目标识别参数还包括所述待识别目标的姿态特征参数；所述将所述目标识别参数与预存的目标认证参数进行比对，生成目标识别结果的步骤，包括：
    确定所述生物特征参数与预存的生物特征参数相符且所述姿态特征参数与预存的姿态特征参数相符，所述目标识别结果为目标识别成功。
  7. 根据权利要求1所述的目标识别方法，其中，所述方法还包括：
    从所述第一光学图像中提取包含第二识别标志物的第二标志物图像，根据所述第一深度图像对所述第二标志物图像中第二识别标志物的特征参数进行校正，其中，所述第二识别标志物包括所述待识别目标相关联的第一物品或与场景相关联的第二物品；
    基于校正后的所述第二标志物图像，提取所述第二识别标志物的目标识别参数；
    基于所述第一识别标志物的目标识别参数和所述第二识别标志物的目标参数，与预存的目标认证参数进行比对，生成目标识别结果。
  8. 根据权利要求7所述的目标识别方法，其中，所述第二识别标志物的目标识别参数为实物特征参数和姿态特征参数。
  9. 根据权利要求1所述的目标识别方法，其中，在所述从所述第一光学图像中提取第一识别标志物的步骤之前，所述方法还包括：
    对所述第一光学图像和所述第一深度图像进行预处理，所述预处理至少包括：所述第一光学图像和所述第一深度图像的图像融合以及去噪处理。
  10. 一种计算机可读存储介质，其上存储有计算机程序，其中，该计算机程序被处理器执行时实现权利要求1至9中任一项所述的目标识别方法。
  11. 一种电子设备，其中，包括：处理器及存储器；
    所述存储器存储有计算机程序，所述处理器被配置成执行所述存储器存储的计算机程序，以使所述电子设备执行如权利要求1至9中任一项所述的目标识别方法。
  12. 一种智能穿戴设备，其中，所述智能穿戴设备包括：
    光学摄像头，被配置成采集包含测试目标的光学图像；
    深度摄像头，被配置成采集包含测试目标的深度图像；
    处理器，所述处理器与所述光学摄像头、所述深度摄像头可通信地耦合；以及
    存储器，所述存储器被配置成存储指令，当所述存储指令被所述一个或多个处理器执行时，使所述处理器执行如权利要求1至9中任一项所述的目标识别方法。
PCT/CN2022/143573 2022-08-29 2022-12-29 目标识别方法、存储介质及设备 WO2024045454A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211047463.4A CN115471416A (zh) 2022-08-29 2022-08-29 目标识别方法、存储介质及设备
CN202211047463.4 2022-08-29

Publications (1)

Publication Number Publication Date
WO2024045454A1 true WO2024045454A1 (zh) 2024-03-07

Family

ID=84369225

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/143573 WO2024045454A1 (zh) 2022-08-29 2022-12-29 目标识别方法、存储介质及设备

Country Status (2)

Country Link
CN (1) CN115471416A (zh)
WO (1) WO2024045454A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115471416A (zh) * 2022-08-29 2022-12-13 湖北星纪时代科技有限公司 目标识别方法、存储介质及设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108764180A (zh) * 2018-05-31 2018-11-06 Oppo广东移动通信有限公司 人脸识别方法、装置、电子设备及可读存储介质
CN109241890A (zh) * 2018-08-24 2019-01-18 北京字节跳动网络技术有限公司 面部图像校正方法、装置及存储介质
JP6631736B1 (ja) * 2019-03-07 2020-01-15 富士通クライアントコンピューティング株式会社 生体認証装置、生体認証システムおよびプログラム
CN111325691A (zh) * 2020-02-20 2020-06-23 Oppo广东移动通信有限公司 图像校正方法、装置、电子设备和计算机可读存储介质
CN113420700A (zh) * 2021-07-02 2021-09-21 支付宝(杭州)信息技术有限公司 手掌生物特征采集装置及方法
CN115471416A (zh) * 2022-08-29 2022-12-13 湖北星纪时代科技有限公司 目标识别方法、存储介质及设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108764180A (zh) * 2018-05-31 2018-11-06 Oppo广东移动通信有限公司 人脸识别方法、装置、电子设备及可读存储介质
CN109241890A (zh) * 2018-08-24 2019-01-18 北京字节跳动网络技术有限公司 面部图像校正方法、装置及存储介质
JP6631736B1 (ja) * 2019-03-07 2020-01-15 富士通クライアントコンピューティング株式会社 生体認証装置、生体認証システムおよびプログラム
CN111325691A (zh) * 2020-02-20 2020-06-23 Oppo广东移动通信有限公司 图像校正方法、装置、电子设备和计算机可读存储介质
CN113420700A (zh) * 2021-07-02 2021-09-21 支付宝(杭州)信息技术有限公司 手掌生物特征采集装置及方法
CN115471416A (zh) * 2022-08-29 2022-12-13 湖北星纪时代科技有限公司 目标识别方法、存储介质及设备

Also Published As

Publication number Publication date
CN115471416A (zh) 2022-12-13

Similar Documents

Publication Publication Date Title
US10043308B2 (en) Image processing method and apparatus for three-dimensional reconstruction
Memo et al. Head-mounted gesture controlled interface for human-computer interaction
KR102292028B1 (ko) 제스처 인식 방법, 장치, 전자 기기 및 저장 매체
Hagbi et al. Shape recognition and pose estimation for mobile augmented reality
Han et al. Enhanced computer vision with microsoft kinect sensor: A review
WO2018177337A1 (zh) 手部三维数据确定方法、装置及电子设备
EP3341851B1 (en) Gesture based annotations
Turk et al. Perceptual interfaces
CN110322500A (zh) 即时定位与地图构建的优化方法及装置、介质和电子设备
Feng et al. Features extraction from hand images based on new detection operators
Malassiotis et al. Personal authentication using 3-D finger geometry
WO2022166243A1 (zh) 捏合手势检测识别方法、装置及系统
WO2019214442A1 (zh) 一种设备控制方法、装置、控制设备及存储介质
WO2024045454A1 (zh) 目标识别方法、存储介质及设备
Cai et al. Visual focus of attention estimation using eye center localization
Pires et al. Unwrapping the eye for visible-spectrum gaze tracking on wearable devices
Poon et al. Occlusion-robust bimanual gesture recognition by fusing multi-views
Dondi et al. Development of gesture‐based human–computer interaction applications by fusion of depth and colour video streams
Ding et al. Designs of human–robot interaction using depth sensor-based hand gesture communication for smart material-handling robot operations
Augustauskas et al. Robust hand detection using arm segmentation from depth data and static palm gesture recognition
Kurdthongmee et al. A yolo detector providing fast and accurate pupil center estimation using regions surrounding a pupil
Yang et al. Audio–visual perception‐based multimodal HCI
TW201248456A (en) Identifying contacts and contact attributes in touch sensor data using spatial and temporal features
Chou et al. A robust real-time facial alignment system with facial landmarks detection and rectification for multimedia applications
Lüsi et al. Human head pose estimation on SASE database using random hough regression forests

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22957263

Country of ref document: EP

Kind code of ref document: A1